Generative AI guidelines for administrative use

Background and Purpose

Generative Artificial Intelligence (AI) is a type of artificial intelligence technology that can generate content including audio, code, images, text, music and videos based on data it's been trained on - at the request of a user. UBC acknowledges that Generative AI tools have transformative potential to enable time saving efficiencies and foster innovation and creativity.

However, the use of these tools also entails significant risk. The potential for unintended consequences including misinformation, copyright violation, unethical manipulation, or violation of personal data and privacy, are significant. The use of these tools can also inadvertently contribute to amplifying biases, promoting unfair outcomes, and potentially undermining the trust our stakeholders place in us.

It is important to recognize the limitations of Generative AI and ensure that all activities involving the use of Generative AI tools align with UBC's policies and guidelines, maintaining a balance between innovation and responsible use.

When using these tools for teaching and learning or research purposes, please refer to the guidance provided by the appropriate sources such as https://academicintegrity.ubc.ca, the UBC Library at https://guides.library.ubc.ca/GenAI/teaching-learning-research or CTLT at https://ai.ctlt.ubc.ca

The purpose of this document is to provide guidance on the use of generative AI tools for administrative purposes in a manner consistent with UBC policies and values and applies to faculty, staff or other affiliates employed by and/or conducting business on behalf of UBC.

 

Considerations

When using Generative AI tools in your work at UBC, consider the following:

Reliability

There are legitimate concerns about the use and reliability of AI such that UBC staff and faculty must seek to always strike the balance between opportunity for improvement and automation with managing and mitigating risks to the university and its community.

Internal biases
Generative AI is built on the gender, racial and other biases of the internet and society in general and its outputs can carry those same biases which may result in inappropriate differential treatment and serious consequences for groups of individuals and/or for their human rights.

Copyright infringement
Models may be trained on intellectual assets which are copyrighted and may inadvertently generate content that infringes upon existing intellectual property.

Weaponization
The extent to which Generative AI can be weaponized by threat actors' malicious exploits is unknown and there is real and significant risk of increased cyber and fraud attacks, such as those that use deep fakes for social engineering of personnel.

Data privacy
There are currently no verifiable data privacy assurances nor protections regarding confidential enterprise information associated with many of these tools. Users should ensure that any data or queries they enter into ChatGPT and other tools will not contain any personal or confidential information. Assume inputs will become public information.

Sustainability
UBC is committed to principles of sustainability. Generative AI uses significant amounts of electricity. Users should consider choosing solutions and/or vendors that limit and/or reduce power consumption and leverage high-quality renewable energy to mitigate the impact on your sustainability goals.

Nascent technology
Generative AI is an evolving technology with unanswered questions remaining about its reliability, accuracy, and best-use. As such, this set of guidelines and any associated procedure must be revisited and maintained to ensure compliance with current and emerging regulatory standards and government advice.

 

Requirements for using Generative AI in your work at UBC

Administrative Approval:
If you are required to accept the terms of use of an AI tool, and will be using that tool for UBC business or teaching purposes, even if the tool is free, you are entering into a “Clickthrough Agreement” under the UBC Board of Governor's Signing Resolution #26, which states that your Administrative Head of Unit must have provided you written authority to execute that Clickthrough Agreement, and that Administrative Head of Unit is responsible for: (a) any liability arising from such Clickthrough Agreement; (b) maintaining a copy of such Clickthrough Agreement; and (c) providing an annual accounting of Clickthrough Agreements to UBC Supply Management.

Appropriate Use:
Ensure it is appropriate to use generative AI tools in the course of your work by confirming with your supervisor the way you intend to use these tools, and any controls you will apply such as ensuring no confidential UBC data is provided to these tools and that appropriate quality assurance is followed.

Understanding Generative AI:
Before utilizing generative AI, make sure you first learn about ChatGPT or other tools' strengths, and limitations. Read all terms of use to ensure you understand privacy, security, and copyright implications. Generative AI is a tool to assist in our tasks, not a replacement for human creativity and judgment. UBC's LinkedIn learning offers an array of free courses on how to best use generative AI.

Maintain Accuracy and Credibility:
While AI can generate impressive content, it is important to review and revise the produced material. Ensure that the output aligns with UBC's institutional values, and ensure the content and references are factual.

Ensure Relevance:
AI does not understand context the same way humans do. Therefore, always ensure the content generated is relevant to the topic and the intended audience. As most tools are still undergoing development, there is a risk that generative AI models can be deceived or manipulated.

Promote Ethical Use:
Generative AI should never be used to create misleading or false information. Use this tool responsibly and ethically, always upholding UBC's values.

Privacy and Security:
Ensure the use of generative AI complies with other UBC Policies such as SC 14: Acceptable Use and Security of UBC Electronic Information and Systems, including the Information Security Standards, and aligns with UBC's data privacy requirements for access and security. If using the tool for specific purposes or projects that involve the collection or use of personal information, a Privacy Impact Assessment is required. Otherwise, never feed it sensitive, confidential, or personal information.

If non-UBC systems are being used for UBC business, ensure compliance with the ISS: SC14 - 4.1 To maintain the security of UBC Electronic Information, Users intending to conduct University Business using systems other than UBC Systems must do so in accordance with the Information Security Standards.

Check the tools settings and disable any history and training features. Some generative AI tools, including ChatGPT, train themselves using your chat history. To maintain as much privacy as possible, it is good practice to clear your chat history immediately after using the tool.

Maintain Human Touch:
While AI can aid us significantly in generating content, the final communication should have a human touch, keeping UBC's tone intact.

Training and Updates:
Generative AI models get better with time and practice. Regularly update the AI model with feedback and newer versions to maintain the quality and relevance of its output.

Copyright:
Generative AI tools may have varying policies on ownership of generated content. Some tools might grant you full ownership, while others may retain certain rights. Before using a generative AI tool, make sure you read the terms of use, and fully understand how material that has been generated can be used for personal or enterprise use.

Even if the tool's policy is that you own the output, beware that outputs generated by these tools may not be entirely original, as the tool learned from copywritten content that, if used as your own, could result in claims of copyright infringement or plagiarism.

Citations:
When using AI-generated content in public-facing communications, consider whether it may be appropriate to disclose that fact. If significant AI involvement is present, consider using footnotes to acknowledge the assistance of AI tools such as “This article was written with assistance from [insert name of tool and link]”. Most writing and citation formats now include instructions on incorporating AI generated content, so it is essential to adhere to https://guides.library.ubc.ca/GenAI/cite for generative AI use.

If imagery has been created or modified using AI, it can be credited in much the same way as photo credits are done. For example, “Photo illustration: [insert name of tool/UBC]'.

Keep Learning and Adapting:
Generative AI is a rapidly evolving field. Stay updated on the latest advancements, use cases, and best practices in AI to harness its potential to the fullest.

Reach Out for Help:
When in doubt, do not hesitate to ask for help or further clarification on the use of generative AI. We are all learning, and open dialogue will ensure we make the most of this innovative tool.

Appendix A: Related Policies & Guidelines

Academic Integrity & Research Ethics

Protection of Privacy and Privacy Risk Assessment

Appendix B: Resources